I just got done doing a very extensive CFD model versus full scale moving ground plane tunnel test on the same ground effects/tunnel car. Probably first time done on this continent-and still more than many F1 teams do. Expensive-yes-but the results (not for general disclosure-sorry) are very interesting and show a lot of things you do not get with just CFD. A lot of things that most CFD engineers and software are not aware of (as they usually have no or little actual wind tunnel experience with all the types of cars that are out there.) navier-stokes has some fundamental issues with cars running on the ground. And we have the proof now.
Hopefully your wind tunnel is better calibrated than the one Ferrari have been using on the F150th Italia.
My company does a lot of CFD in the motorsports industry. A lot. We are quite aware of all you mention. Aerospace use is way different animal. In some ways easier--I have done a bunch of aero CFD (No ground, no rotating wheels,).
yea-Virgin, Aston Martin and others. Issue is putting all of your eggs in the CFD basket. I am not anti CFD-far from it. My issue is that there is a growing (and wrong) assumption that CFD will give the "typical" racer a quick and easy and cheap way of getting aero data. It just ain't so. (Yet)-give it 10 years to mature but even then it still will not be a "press a button" and get a car design.
I just got done doing a very extensive CFD model versus full scale moving ground plane tunnel test on the same ground effects/tunnel car. Probably first time done on this continent-and still more than many F1 teams do. Expensive-yes-but the results (not for general disclosure-sorry) are very interesting and show a lot of things you do not get with just CFD. A lot of things that most CFD engineers and software are not aware of (as they usually have no or little actual wind tunnel experience with all the types of cars that are out there.) navier-stokes has some fundamental issues with cars running on the ground. And we have the proof now.
What is your company called or who do you work for, mate?
Mr. Gocarts:
Like they say, if it were easy, everybody would do it! I think that applies to both CFD and experiments. Knowledge and care is needed to do either one successfully. As a mechanical engineering fluid dynamics experimentalist who has dabbled in CFD, my feeling is that there are more pitfalls for the novice in CFD. One has to be sure of grid independence and boundary condition sensitivity, and somehow validate for computations of an unknown flow. One has to choose an appropriate turbulence model and probably an appropriate wall function. It is well known that simple turbulence models like k-epsilon have some problems with accuracy doing separated flows, which are common on our cars. I don't think one can just say that reduced accuracy is acceptable for showing trends, for the reduced accuracy in predicting things like transition and separated flow behavior can disguise or reverse trends. This is not even considering the complexities of rotating wheels and motion relative to the ground with very small clearances.
I don't believe that LES is the solution many thought it would be ten years ago, as it has been found to have problems with wall boundary conditions. A hybrid approach, Discrete Eddy Simulation, combining zones with LES to near wall zones using RANS models now seems to be the trend.
What percentage improvements are you looking for? Are these within the uncertainty of the method you are applying? This applies to both wind tunnel testing and CFD. I think amateur wind tunnel testing with small models is particularly problematic for lots of the reasons you mention and more. The F1 engineers are looking for very small improvements. It is amazing they do as well as they do given the complexity of the flows. I always wonder about how well the various vanes work when the cars are in the wake of another car.
Knowledge and care is needed to do either one successfully.
Absolutely, but to get to that point you have to learn - which means trying it, making mistakes and then learning some more.
Quote:
Originally Posted by Frank C
I don't think one can just say that reduced accuracy is acceptable for showing trends, for the reduced accuracy in predicting things like transition and separated flow behavior can disguise or reverse trends. This is not even considering the complexities of rotating wheels and motion relative to the ground with very small clearances.
CFD isn't perfect, but that doesn't make it useless. There are all manner of reduced accuracy (less than perfect) methods that engineers, such as yourself, have used throughout time to build structures, cars, airplanes, etc. It's the engineer's job to decide which is the appropriate tool and how to use it. Just because a model isn't perfect under certain circumstances doesn't mean it's useless. The better engineers will make better decisions, succeed, and win races.
Rotating wheels and moving ground planes are a basic feature supported by most CFD software. RANS CFD can adequately model both. Just because the setup looks complicated doesn't mean CFD can't model it.
Predicting transition isn't a strength of RANS CFD, yet in trend analysis it isn't that important. Sure it's important if you want to nail the drag coefficient, but in racing drag is usually secondary to downforce anyway.
Quote:
Originally Posted by Frank C
I don't believe that LES is the solution many thought it would be ten years ago, as it has been found to have problems with wall boundary conditions. A hybrid approach, Discrete Eddy Simulation, combining zones with LES to near wall zones using RANS models now seems to be the trend.
More efficient in terms of run time and memory usage maybe, but with the right resources (i.e., massive) LES works fine, just ask the folks at TotalSim http://totalsim.us/
Unless you have a duff engine-then most issues with a slow car in F1 or prototype are aerodynamic related. The influence on aero to performance is so critical nowadays that a less than optimal aero package means car is pretty well uncompetitive. You have to mess up real big in other areas to make a difference.
On another note: 10 million cell size for an open wheel half/model? A bit sparse-I typically am up around 40 million cell size for a prototype car half model. And that is with coarse refinement in too many areas. And that takes a Linux cluster of 40 computers a day to run.
I managed a CFD wind tunnel correlation study about 6 years ago. It took 2 years to complete and about 10 man years of work. Once we understood what it took to get decent correlation the CFD work became much more usefull and "generally" provided directionally correct results.
Unless you have a duff engine-then most issues with a slow car in F1 or prototype are aerodynamic related. The influence on aero to performance is so critical nowadays that a less than optimal aero package means car is pretty well uncompetitive. You have to mess up real big in other areas to make a difference.
On another note: 10 million cell size for an open wheel half/model? A bit sparse-I typically am up around 40 million cell size for a prototype car half model. And that is with coarse refinement in too many areas. And that takes a Linux cluster of 40 computers a day to run.
And I am waay behind the F1 boys....
10million is a minimum, and by full car I mean a full vehicle simulation, but a symmetric half model. This is a minimum to capture trends (not absolute values) - and it assumes you understand very well the level of geometry details required and your meshing methods are very well developed/extremely efficient.
Le Mans cars or F1 cars, etc. are of course bigger cases as there are much more complex features to capture.
Last edited by aerogt3; May 11th, 2012 at 5:02 AM.
A 10 million cell model of what we are doing would just give us bad results. We know as we have tried. But then we have full rad models, squished tires, exhaust energy, etc. I understand the appeal and value of smaller models to give trends but one must be very, very careful with interpreting the results. We have seen a small change in model complexity give a (correct and accurate) 5-6% change in balance but overall L/D stays the same. (Close enough). My issue is that people like to see CFD as a wonder tool that the local, regional race guy/mfg can use to solve all of his aero issues.
Improper CFD (which is a lot) is worse than no CFD. And how many people (Like Jay Novak) go thru the process of correlation with tunnel or real world. And yes I know the pitfalls of tunnel testing and real life testing quite well. But it is because I know the pitfalls that I can have confidence in what I get from those tests. With CFD there are still so many unknowns.
This is not a knock on CFD-we use it a lot and use it well-but we realize that it is not a simple tool for the "average" guy, team or mfg to use. It is a lot harder and a lot trickier than advertised. And the hardware costs in CPU is pretty darn high, if you want reasonable analysis times.
Check out what Totalsim in Ohio have-they are probably the best equipped guys in US to do contract CFD. Their model size runs around 50m cells and have a cluster that takes about 12 hours a run. You can do less-but you will not get anywhere as accurate a set of results. Just the way it is.
This is one of the coolest threads I've read in awhile. But, I think for my old tractor powered Formula Ford I'll just tape a bunch of wool tufts on it take pictures.
__________________
Man will race anything. It's in his blood. His Soul. He must.
Love hearing about other peoples CFD experiences! We run sims with simplified bluff bodies (half cut sims) at ~8-10m cells for a tiny FSAE car, which we later validate with wind tunnel and real-life testing/data aq. the best thing I can say is a simple simulation can point you in the right direction to getting the car set up... if you have nowhere to start from, well, how awful would that be?
Ive wanted to try open foam and caedium and see how they correlate to other packages, it certainly would be interesting.
the best thing I can say is a simple simulation can point you in the right direction to getting the car set up... if you have nowhere to start from, well, how awful would that be?
Well, poorly executed simulation can point you in the WRONG direction!! This risk is small when you're talking about setting up a car, because really you're just wasting the few laps it takes to see it in the laptimes. But on a design, this can be very expensive! Generally, the bigger the changes you make, the less likely it is to misguide you (or alternatively, the smaller the changes you want to see, the more particular you have to be about your methods.)
And in upper level motorsports even very good simulation and testing can (and do) point you in the wrong direction! I work on the racing programmes of an OEM where at least 30-40% of what CFD says will work just doesn't when it hits the track, and the resources involved are quite good. Granted these are really small changes - even the wind tunnel misses it. I do CFD for a living so I'm very pro-CFD; just want to warn that it's not always right of course.
Well, poorly executed simulation can point you in the WRONG direction!! This risk is small when you're talking about setting up a car, because really you're just wasting the few laps it takes to see it in the laptimes. But on a design, this can be very expensive! Generally, the bigger the changes you make, the less likely it is to misguide you (or alternatively, the smaller the changes you want to see, the more particular you have to be about your methods.)
And in upper level motorsports even very good simulation and testing can (and do) point you in the wrong direction! I work on the racing programmes of an OEM where at least 30-40% of what CFD says will work just doesn't when it hits the track, and the resources involved are quite good. Granted these are really small changes - even the wind tunnel misses it. I do CFD for a living so I'm very pro-CFD; just want to warn that it's not always right of course.
Except my simulations
Right there with you! If you cannot verify your simulations in real life (a from scratch design process or something), I surely see that being a problem. Luckily, we had something to validate initially, and tweak our sims from there. We have some tricks with airfoil design, but those I cant share.
Right there with you! If you cannot verify your simulations in real life (a from scratch design process or something), I surely see that being a problem. Luckily, we had something to validate initially, and tweak our sims from there. We have some tricks with airfoil design, but those I cant share.
Been trolling the forum for a while now but signed up only recently. Hey everyone.
This is an excellent discussion, many thanks to the many of you who have weighed in with your experiences.
Raneff, can you elaborate on why you can't share? I'd be interested in hearing just how far FSAE has come since my college days.
General Formula Car Discussion
